procedure is not described and it is not possible to determine whether allocation was concealed. It is questionable whether the outcome variables of nutritional intake fall within the field defined in our protocol (mood problems, affective behaviour and cognitive decay), but both reviewers found that they do belong to the intended overall problem of care for the health and quality of life for persons suffering from dementia. However, data are not reported in sufficient detail to enter into further analysis. Both reviewers were in doubt as to whether the outcome - nutritional intake - should be considered as belonging to the cluster of outcomes defined in the protocol. It was finally decided that it does, for material reasons. Nutritional intake can be seen as a surrogate measure of the behavioural and mood problems associated with dementia, and the authors explicitly characterize it as such in their introduction: “Persons with chronic organic brain syndrome who are confused and disoriented often wander from the table and leave the food untouched...” Remington 2002: RCT assessing a short-term effect of calming music and hand massage by comparing four treatment arms (calming music, hand massage, simultaneous calming music and Massage and touch for dementia (Review) 5 Copyright © 2008 The Cochrane Collaboration. Published by John Wiley & Sons, Ltd. hand massage, no intervention). Sixty-eight nursing home residents with dementia (medical records contained a diagnosis of Alzheimer’s disease, multi-infarct dementia or senile dementia) were randomly allocated using a sealed envelope procedure to four groups with 17 persons in each group. The intervention lasted for 10 minutes, and was given to each patient once. Treatment effect on ’agitation level’ was evaluated by using a modified version of the Cohen-Mansfield Agitation Inventory (CMAI) administered by trained research assistants who were blinded to treatment allocation when possible (obviously this is not possible for measurements during the intervention). This study met all of the quality requirements except that the description of the procedure indicates that patients could have been excluded after allocation (if they had a CMAI score of 0 at baseline). However, the authors have indicated that in fact no such exclusions happened. Furthermore, they kindly supplied additional data which allowed us to enter the study into RevMan for meta-analysis. However, in the absence of any other comparable data, this adds little to the data analysis already given in the published paper. Excluded studies Giasson 1999: RCT assessing the short-term effect of a ’therapeutic touch’ intervention on patient ’discomfort’ measured by a validated behaviour observation scale. Twenty-seven nursing home residents diagnosed as suffering from dementia of the Alzheimer type, stage 5 to 7, were randomized to an intervention group (n = 16) and a control group (n = 11) which received ’simple presence’ as a control intervention. Interventions lasted 10 minutes and were given once per day for five consecutive days. Observers and evaluators were blinded to group allocation, and observation took place three times: immediately before the first intervention, and immediately after interventions 3 and 5. The authors report a decrease in discomfort levels which was significantly greater in the treatment than in the control group. However, the report lacks a clear summary of means and standard deviations (or other measures of variation) for effect data. Furthermore, the allocation procedure is not specified further than the indication that it involved randomization (’facon aleatoire’). In particular, there is no indication of concealed allocation. No explanation is given for the large difference in group sizes, and there is no information on compliance and dropouts. Hence it seems likely that there were either irregularities in the allocation procedure or very large dropout rates. Scherder 1995; Scherder 1995a; Scherder 1998: Three reports of a series of very similar controlled trials - according to correspondence with the authors, the latter two are in fact reports from the same trial, but using different outcome measures. The trials assess the long-term effects of a combined intervention of transcutaneous electrical nerve stimulation (TENS) and back massage (Scherder 1995) or back massage alone (Scherder 1995a; Scherder 1998) on cognitive function as measured by a battery of standard tests (Scherder 1995a), affective behaviour as measured by another battery of tests (Scherder 1998) or a combination (Scherder 1995). In all cases, the control intervention was sham TENS. In all of the group’s studies, 16 nursing home residents with a diagnosis of dementia of probable Alzheimer’s type were allocated to treatment and control groups, and interventions were given for 30 minutes per day for six weeks. Evaluators were blinded to group allocation, and pre-, post- and delayed scores (measured after a ’washout’ period of six weeks) were recorded. The authors report statistically significant improvements between pre- and post-values in some of the several scales and subscales used, but only small and insignificant differences between pre- and delayed values. Only one of the reports (Scherder 1998) specifies that allocation was random, but according to correspondence